Grammarly CEO on trust, AI, and the future of knowledge work

Executive overview

AI is flooding workplaces with cheap content, but hasn't yet answered who will read it all. Rahul Roy-Chowdhury argues the real opportunity is not content generation but workflow reimagination — synthesising information, preserving individual voice, and keeping humans in control of their thinking.

Grammarly's response to LLM competition rests on 15 years of compounding advantages: deep user trust, privacy infrastructure, and a presence across every tool where communication happens.

The person who knows how to use AI will take the job from the person who doesn't.

How Grammarly thinks about competition

• New LLM-powered rivals (ChatGPT, Copilot, Pi) are welcomed, not feared — competitive pressure sharpens user focus.
• Historical scope was limited to the revision phase of communication by technology constraints, not by choice.
• LLMs now allow Grammarly to span the full cycle: conception, composition, revision, and comprehension.
• Cross-tool context is the next frontier — pulling a Salesforce update into a Slack reply at the right moment.
• The goal is removing drudgery and freeing attention for higher-level thinking.

The TRUE framework (trust, responsibility, user control, empathy)

• Built from existing practices; named to give decision-making a north star in a fast-moving landscape.
• Every user — free or paid — can opt out of having their data used to train Grammarly's models.
• Seismograph: an open-sourced sensitive-text classifier that detects hate speech and problematic outputs; can suppress or re-prompt the LLM before a user sees harmful content.
• Grammarly's decade-plus investment in security and privacy pays larger dividends now than when it was made.
• The cheese-glue incident illustrates how low-quality training data creates industry-wide trust damage; quality annotation and output evaluation are prosaic but essential mitigations.

On open source

• Linux beating Windows in the data-centre battle is the template: openness scales better than closed systems.
• Transparency is the surest historical mechanism for hardening tools against misuse.
• Grammarly currently mixes open and closed models; the long-term bet is on open source as the foundation for trustworthy AI.
• Open source also allows best practices — like Seismograph — to propagate across the industry.

AI and the future of work

• Dropping content-creation cost to near zero increases supply dramatically; without better synthesis tools, the information overload gets worse, not better.
• The ATM precedent: automation reduced routine transactions but multiplied bank branches and teller headcount, shifting work to higher-value tasks.
• "AI takes your job" framing misses the point — it is the skill of using AI that becomes the differentiator.
• Workers are not passive observers; active adoption of AI tools is a choice available to everyone now.
• Grammarly's vision: reimagined workflows where AI synthesises, surfaces insights, and handles drudgery — leaving judgment and creativity to the human.

Voice, agency, and the limits of automation

• Writing is thinking; outsourcing writing risks outsourcing critical thought.
• The way someone communicates is core to their identity — a tool that flattens voice to a generic lowest common denominator is counterproductive.
• Grammarly's role is companion, not replacement: superpowers for the user's own voice, not a substitute for it.
• User control and agency are non-negotiable design constraints, not optional features.

Values in practice

• Grammarly pulled out of Russia and Belarus immediately after the Ukraine invasion — founders are Ukrainian and staff remain based in Kyiv.
• The decision was framed as showing solidarity with colleagues in harm's way, not as a political statement.
• The company engages with hundreds of universities navigating AI and academic integrity — including an AI-citation feature that logs how students actually used Grammarly prompts so educators can assess genuine engagement.
• Mark Francis Long (dyslexia advocate, "I Am Lex") is cited as a motivating example of Grammarly's access impact; similar stories arrive weekly.
• Virtue signalling on every social issue is explicitly rejected; the bar for a public stance is clear values alignment, not visibility.